Enterprise-level AI Agent for Plan-and-Execute + Function Calling

tools = [

	{
		"type": "function",
		"function": {
			"name": "Search Database",
			"description": "Retrieve profile from database",
			"parameters":{
				"type": "object",
				"properties": {
					"name": {"type": "string"}
				},
				"required": ["name"]
			}
		}
	},
	{
		"type": "function",
		"function": {
			"name": "Send Mail",
			"description": "Send Mail to Gmail",
			"parameters":{
				"to": {"type": "string"},
				"subject": {"type": "string"},
				"body": {"type": "string"}
			}
		}
	},
	{
		"type": "function",
		"function": {
			"name": "Get Weather",
			"description": "Get Weather from location",
			"parameters":{
				"type": "object",
				"properties": {
					"location": {"type": "string"}
				},
				"required": ["location"]
			}
		}
	},
......

]

prompt = ‘
	You are a task planner in a multi-step AI system.
	User query: {Query}
	Your job:
	1, Break down the user query into executable steps.
	2, Each step must be atomic and tool-executable.
	3, Steps should be ordered logically.
	4, Do NOT execute anything.
	5, Only return JSON.
’

import json
 
# 1️⃣ Let the Planner generate the plan
planner_response = client.chat.completions.create(
    model="gpt-4o",
    messages=[
        {"role": "system", "content": planner_prompt},
        {"role": "user", "content": user_query}
    ]
)
 
plans = json.loads(planner_response.choices[0].message.content)["plans"]
 
# 2️⃣ Execute step-by-step
context_memory = {}
 
for step in plans:
    tool_name = step["tool"]
    args = step["arguments"]
 
    result = execute_tool(tool_name, args, context_memory)
 
    # Add the result to context
    context_memory[f"step_{step['step_id']}"] = result

def execute_tool(tool_name, args, memory):
 
    if tool_name == "Search Database":
        return search_Database(args["name"])
 
    elif tool_name == "Send Mail":
        
        return send_Mail(args["to"],args["subject"],args["body"])
 
    elif tool_name == "Get Weather":
        return get_Weather(args["location"])
 
    else:
        raise Exception("Unknown tool")

Advantages:

1. Task Decomposition: Breaking down large tasks into smaller ones enables effective management and resolution of complex tasks.
2. Detailed Guidance: Providing detailed instructions improves the quality and accuracy of reasoning steps.
3. Adaptability: Can be adjusted for different types of tasks and performs well on various complex problems.

Disadvantages:

1. The Planner Prompt generates the plan all at once; steps may be missed, unordered, and intermediate states are not perceivable.
2. The Execute loop is too mechanical, lacking conditional jumps, branching, dynamic decisions, and failure retries.
3. The Planner does not know the execution results, such as whether they are empty or not.

Suitable Scenarios:

1. Fixed steps and stable workflows
2. No need for dynamic decision-making or branching logic

Unsuitable Scenarios:

1. Decision-driven tasks
2. Conditional jumps
3. Data-driven reasoning
4. Long-chain tasks